US2025371691A1PendingUtilityA1
Device for diagnosing defect using model based on continual learning and method therefor
Est. expiryMay 28, 2044(~17.8 yrs left)· nominal 20-yr term from priority
G06T 2207/30168G06T 2207/20081G06T 2207/10116G06T 7/0002G06N 3/096G06T 2207/20084G01N 23/04G06N 20/00G06V 10/82G06V 10/774
59
PatentIndex Score
0
Cited by
0
References
0
Claims
Abstract
Proposed are a device for diagnosing a defect using a model based on continual learning and a method therefor, and the method for diagnosing the defect includes a step of loading, by a training unit, a buffer training data set, which is training data selected according to a degree of influence on prediction performance of a current model from among a past training data set in the continual learning, a step of training the current model using the buffer training data set and a current training data set, and a step of detecting, by a detection unit, the defect in a radiographic image using the current model when the radiographic image is input.
Claims
exact text as granted — not AI-modifiedWhat is claimed is:
1 . A method for diagnosing a defect, the method comprising:
training, by a training unit, a current model using a buffer training data set previously stored in a buffer and a current training data set; and detecting, by a detection unit, the defect in a radiographic image using the current model when the radiographic image is input.
2 . The method for claim 1 , wherein the training of the current model comprises:
loading, by the training unit, the buffer training data set, which is training data selected according to a degree of influence on prediction performance of the current model from among a past training data set in continual learning; and training the current model using the buffer training data set and the current training data set.
3 . The method for claim 1 , further comprising:
calculating, by a switching unit, after the training of the current model, a degree of change from models trained in at least two previous training rounds of the current model in the continual learning to the current model; determining, by the switching unit, whether the degree of change is greater than or equal to a reference value; determining, by the switching unit, an update to the buffer training data set when the degree of change is greater than or equal to the reference value; calculating, by an update unit, a degree of influence of each of data points in the current training data set on prediction performance of a future model when the future model to be trained in a next training round of the current model is trained using the current training data set in the continual learning; and updating, by the update unit, the buffer training data set by selecting a predefined number of the data points in order of higher to lower degrees of influence.
4 . The method for claim 3 , wherein the calculating of the degree of influence comprises:
calculating, by the update unit, a plasticity score representing a first probability that a first prediction value of the current model and a second prediction value of the future model are different from each other for the data points in the current training data set; calculating, by the update unit, a stability score representing a second probability that the first prediction value of the current model and a third prediction value of a first past model trained in a previous training round of the current model, are different from each other for the data points in the current training data set; and determining, by the update unit, a weighted average of the plasticity score and the stability score as the degree of influence.
5 . The method for claim 3 , wherein calculating a stability score is performed according to Equation
p
(
y
^
t
-
1
≠
y
^
t
❘
"\[LeftBracketingBar]"
x
t
)
=
1
-
p
(
y
^
n
t
-
1
❘
"\[LeftBracketingBar]"
y
^
n
t
,
x
n
t
)
,
wherein p(ŷ t−1 ≠ŷ t |x t ) denotes the stability score,
x
n
t
denotes the data points in the current training data set,
y
^
n
t
-
1
denotes a first prediction value of a first past model, and
y
^
n
t
denotes a second prediction value of the current model.
6 . The method for claim 3 , wherein calculating a plasticity score is performed according to Equation
p
(
y
^
t
+
1
≠
y
^
t
❘
"\[LeftBracketingBar]"
x
t
)
=
1
-
p
(
y
^
n
t
+
1
❘
"\[LeftBracketingBar]"
y
^
n
t
,
x
n
t
)
,
wherein p(ŷ t+1 ≠ŷ t |x t ) denotes the plasticity score,
x
n
t
denotes the data points in the current training data set,
y
^
n
t
+
1
denotes a first prediction value of the future model, and
y
^
n
t
denotes a second prediction value of the current model.
7 . The method for claim 6 , wherein the calculating of the plasticity score derives the prediction value of the future model according to Equation
p
(
y
^
t
+
1
❘
"\[LeftBracketingBar]"
x
t
)
=
f
(
GP
(
θ
t
)
,
x
)
,
wherein
y
^
n
t
+
1
denotes the first prediction value of the future model,
x denotes the data points in the current training data set, and
GP(θ t ) denotes a gradient vector predicted by the future model through a gradient prediction model.
8 . The method for claim 3 , wherein the calculating of the degree of influence is performed according to Equation
S
i
=
λ
·
Plasticity
i
+
(
1
-
λ
)
·
Stability
i
,
wherein S denotes the degree of influence,
i denotes an index of the data points in the current training data set,
λ denotes a weight,
Plasticity denotes a plasticity score, and
Stability denotes a stability score.
9 . The method for claim 3 , wherein the calculating of the degree of change derives the degree of change according to a degree of similarity between:
a first gradient vector representing a first change between a first weight vector of the current model and a second weight vector of a first past model trained in the previous training round of the current model, and a second gradient vector representing a second change between the second weight vector of the first past model and a third weight vector of a second past model trained in a previous training round of the first past model.
10 . The method for claim 9 , wherein the degree of similarity is calculated according to Equation
CS
=
g
t
-
1
T
g
t
g
t
-
1
2
g
t
2
,
wherein CS denotes the degree of similarity,
g t denotes the first gradient vector between the first weight vector of the current model and the second weight vector of the first past model,
g t−1 denotes the second gradient vector between the second weight vector of the first past model and the third weight vector of the second past model, and
g
t
-
1
T
denotes a transpose vector of the second gradient vector.
11 . A device for diagnosing a defect, the device comprising:
a training unit configured to train a current model using a buffer training data set previously stored in a buffer and a current training data set; and a detection unit configured to detect the defect in a radiographic image using the current model when the radiographic image is input.
12 . The device of claim 11 , wherein the training unit loads the buffer training data set, which is training data selected according to a degree of influence on prediction performance of the current model from among a past training data set in continual learning, and trains the current model using the buffer training data set and the current training data set.
13 . The device of claim 11 , further comprising:
a switching unit configured to: calculate a degree of change from models trained in at least two previous training rounds of the current model in the continual learning to the current model; determine whether the degree of change is greater than or equal to a reference value; and determine an update to the buffer training data set when the degree of change is greater than or equal to the reference value, and an update unit configured to: calculate a degree of influence of each of data points in the current training data set on prediction performance of a future model when the future model to be trained in a next training round of the current model is trained using the current training data set in the continual learning; and update the buffer training data set by selecting a predefined number of the data points in order of higher to lower degrees of influence.
14 . The device of claim 13 , wherein the update unit calculates a plasticity score representing a first probability that a first prediction value of the current model and a second prediction value of the future model are different from each other for the data points in the current training data set, calculates a stability score representing a second probability that the first prediction value of the current model and a third prediction value of a first past model trained in a previous training round of the current model, are different from each other for the data points in the current training data set, and determines a weighted average of the plasticity score and the stability score as the degree of influence.
15 . The device of claim 13 , wherein the update unit calculates a stability score according to
Equation
p
(
y
^
t
-
1
≠
y
^
t
❘
"\[LeftBracketingBar]"
x
t
)
=
1
-
p
(
y
^
n
t
-
1
❘
"\[LeftBracketingBar]"
y
^
n
t
,
x
n
t
)
,
wherein p(ŷ t−1 ≠ŷ t |x t ) denotes the stability score,
x
n
t
denotes the data points in the current training data set,
y
^
n
t
-
1
denotes a first prediction value of a first past model, and
y
^
n
t
denotes a second prediction value of the current model.
16 . The device of claim 13 , wherein the update unit calculates a plasticity score according to
Equation
p
(
y
^
t
+
1
≠
y
^
t
❘
"\[LeftBracketingBar]"
x
t
)
=
1
-
p
(
y
^
n
t
+
1
❘
"\[LeftBracketingBar]"
y
^
n
t
,
x
n
t
)
,
wherein p(ŷ t+1 ≠ŷ t |x t ) denotes the plasticity score,
x
n
t
denotes the data points in the current training data set,
y
^
n
t
+
1
denotes a first prediction value of the future model, and
y
^
n
t
denotes a second prediction value of the current model.
17 . The device of claim 16 , wherein the update unit derives the prediction value of the future model according to
Equation
p
(
y
^
t
+
1
❘
"\[LeftBracketingBar]"
x
t
)
=
f
(
GP
(
θ
t
)
,
x
)
,
wherein
y
^
n
t
+
1
denotes the first prediction value of the future model,
x denotes the data points in the current training data set, and
GP(θ t ) denotes a gradient vector predicted by the future model through a gradient prediction model.
18 . The device of claim 13 , wherein the update unit calculates the degree of influence according to
Equation
S
i
=
λ
·
Plasticity
i
+
(
1
-
λ
)
·
Stability
i
,
wherein S denotes the degree of influence,
i denotes an index of the data points in the current training data set,
λ denotes a weight,
Plasticity denotes a plasticity score, and
Stability denotes a stability score.
19 . The device of claim 13 , wherein the switching unit derives the degree of change according to the degree of similarity between:
a first gradient vector representing a first change between a first weight vector of the current model and a second weight vector of a first past model trained in the previous training round of the current model, and a second gradient vector representing a second change between the second weight vector of the first past model and a third weight vector of a second past model trained in a previous training round of the first past model.
20 . The device of claim 19 , wherein the switching unit calculates the degree of similarity according to
Equation
CS
=
g
t
-
1
T
g
t
g
t
-
1
2
g
t
2
,
wherein CS denotes the degree of similarity,
g t denotes the first gradient vector between the first weight vector of the current model and the second weight vector of the first past model,
g t−1 denotes the second gradient vector between the second weight vector of the first past model and the third weight vector of the second past model, and
g
t
-
1
T
denotes a transpose vector of the second gradient vector.Join the waitlist — get patent alerts
Track US2025371691A1 — get alerts on status changes and closely related new filings.
We store only your email — no account needed. See our privacy policy.